Optimize LLM cost and latency with prompt caching: cacheable prefixes, cache-control headers, context layout, and cache-aware prompting. Use for high-volume apps.
